|
|
| Research Interests | |
| Nonparametric statistics: minimax and adaptive estimation, oracle inequalities, aggregation, and Le Cam's theory of equivalence of experiments. | |
| Semiparametric statistics: second-order optimality, profiled maximum likelihood estimation, goodness-of-fit testing, sinle-index and multi-index models. | |
| Diffusion processes: stationarity, drift estimation, non-synchronous sampling. | |
| Machine learning: classification, exponential weighting, feature selection. | |
| High dimensional statistics: sparse recovery and dimension reduction. | |
| Computer vision: structure from motion, image denoising, pattern recognition. | |
| Current Projects | |
|
ANR CALLISTO
[ Coordinator: Pascal Monasse | 380 586 |
project homepage ] La précision finale de la géométrie estimée par vision stéréo multi-vues dépend de toutes les étapes de la chaîne algorithmique. Alors que des méthodes précises de localisation de points d'intérêt sont bien connues, les étapes initiales, à savoir les calibrations interne et externe, sont souvent considérées comme correctes, puis ajustées dans une optimisation globale, le « bundle adjustment ». L'énergie à minimiser est loin d'être convexe, dépend de nombreuses variables, et a de fortes chances de présenter de multiples minima. Il semble préférable d'estimer séparément les paramètres internes de caméra et les positions et orientations des vues. Ce projet propose entre autres d'évaluer un modèle non paramétrique de distorsion géométrique de la lentille et de fournir une méthode globale de calibration externe. Les outils de choix sont issus de méthodes statistiques récentes. |
|
|
ANR PARCIMONIE
[ Coordinator: Erwan Le Pennec | 250 000 |
project homepage ] The PARCIMONIE project is dedicated to the study of sparsity based estimators. Our main goal is to derive several novel sparsity based estimators and investigate their properties using various criteria: theoretical ones (minimaxity, maxisets, oracle type inequalities), experimental ones (comparison on simulation, competition on several common targets). Applications of our methods is an important issue as well as a motivation. As a consequence an important part of the project is oriented toward applications. We need to understand the type of representations that can be used, choosing the proper building blocks; we need to implement the estimators. We will also look at real applications in Biology, Astrophysics and Image processing, to tailor our constructions to their specific requirements. It is the common belief of the participants that there should be continuous feedbacks between \(\ \ \bullet\) sparse estimation (how to select atoms?), \(\ \ \bullet\) representations (which atoms/representation to use?), \(\ \ \bullet\) algorithms (how to compute the estimators?) and, \(\ \ \bullet\) applications (which estimator to use?). |
|
| Last update: June 28, 2011 |